Most AI regulation is security theater with paperwork. By late 2026, the standard prescription is still labels, disclosures, and model cards—rules that make AI feel supervised while leaving the actual blast radius untouched.

I’ve watched this pattern outside AI too. Tobacco companies complied with warning labels and kept selling cigarettes. Social platforms added disclosure banners and kept harvesting attention. Compliance becomes a decorative layer that lets the public feel protected without forcing the industry to change.

The missed point is accountability. If an AI system screens tenants, grades students, denies loans, generates medical advice, or makes hiring decisions, the person or company deploying it should carry the legal responsibility. “The model did it” is not an excuse; it is the business model.

Regulating model makers is also backwards. Open-source systems are not the central danger. Powerful companies deploying AI at scale are. Rules should attach to use, scale, and consequences—not to whether a chatbot has a red dot next to its name.

Labels matter, but they are the least important part of the fight.

AI should be regulated like a consequential tool, not marketed like a magical beverage.